Algorithmic Systems and Methods For Ranking Posts in a Social Platform

Castamir

BANNED
Joined
Sep 4, 2019
Messages
1,200
Reaction score
1,055
This is a skeletal statement of how a social media algorithm could work.

Have in mind that this is designed for a small social media where the users are not connected by followers or friendships.

I'm looking to make it better.

If you have any more ideas, please contribute below.

The implementation of this algo makes any post that got any recent engagement (not number, but how recent) to rise to the top.

>>> Example of Recent Activity Based Ranking Algorithm.

Post A gets published at 11:15am

Post B gets published at 11:16am

Post C gets published at 11:17am

In the absence of any engagement (likes and comments) occurring between 11:18am till 11:20am, the page when refreshed looks like this.

#1 Rank - Post C (since it got published recently)

#2 Rank - Post B

#3 Rank - Post A

NOTE: List of actions counted as an activity "Publishing, comments, likes, shares". So, how recent an activity is, is the most important ranking factor for the post that recieved the activity.

>>> Further Implementation of the Algorithm Illustrating the Note Above

Post B gets a like at 11:22am

Post C gets a share at 11:20am

Post D gets published at 11:21am

Post A gets a comment at 11:19am


The ranking will look as follows based on recent activity (See note above)

#1 - Post B
#2 - Post D
#3 - Post C
#4 - Post A


ADVANCED CONDITIONS FOR THE ALGO

As to be expected in such small social platforms, there may be spam and low quality posts or boring posts. These are conditions for what users see on their newsfeed.

1. To avoid spam or post overload, only the first post per a user account made in an hour is selected for inclusion on the home page. All other posts can only be seen on the profile page.

2. All posts by verified accounts will always be on the homepage.

3. Posts liked by verified accounts get to the homepage.

4. Posts older than a week cannot get to the homepage.

5. All posts has an impression/views counter. A post on the homepage with 20 views and zero likes or comments gets removed from the homepage.

6. A post with at least one one censored word does not get to the homepage unless it was posted or liked by a verified account.

7. All posts with a link not contained in the group of whitelisted or approved domains does not get to the homepage.

8. No post from a user account whose content has been reported at least once will get to the homepage for one week.


Please criticize or agree with the conditions I have stated. What more conditions do you think will be suitable?
 
Last edited:
No. 9 - A post may satisfy condition number 5 and go on to get many views and likes. If the likes to view ratio is less than 4%. It gets removed from the newsfeed and will only be accessible from the profile page.
 
If I am to hire developers, what skills do I need to ask for this to be accomplished?
 
The feed should be catered to each user or you will end up showing irrelevant content to everybody. For this you need some form of a recommendation engine using tags. For 'trending' or 'popular' content, try using streaming solutions such as Apache Kafka.
 
The feed should be catered to each user or you will end up showing irrelevant content to everybody. For this you need some form of a recommendation engine using tags. For 'trending' or 'popular' content, try using streaming solutions such as Apache Kafka.
Is there a one stop solution for that?
 
Why don't you ask how does the BHW home page rates the posts displayed?
Would give a good idea if they are willing to share.
Basically, if you look at the algo, you will see it works just like BHW but with a few refinements.
 
Back
Top